OpenResearchAlpha

Interactive Journal of Medical Research · 13 Nov 2026

Effects of a Personalized Retrieval-Augmented Generation Chatbot on Information Needs in Patients With Chronic Kidney Disease: Mixed Methods Randomized Controlled Trial

Hee Jeong Hwangh-index 1, Junghwan Kim, Dongseok Heo, Jeong Yun You, Jeonghwan Lee, Soie Kwon, Kyung Don Yoo, Jung Han Yoon Park, Bongwon Suh, Jung Pyo Lee, Ki Won Leeh-index 7

0
Citations
no database records one yet
0
Citations in 2026
so far
7
Leading author's h-index
Ki Won Lee, in QAI's scholar index
35
References
works it cites, from open citation data

Citations by database

Each database indexes a different set of journals, preprints and books, so they are shown side by side rather than blended.

  • OpenCitations0
  • Crossref0
  • Europe PMC0

Citations per year

The current year is still counting.

  • 2026 so far0

Abstract

Abstract Background Patients receiving dialysis for chronic kidney disease encounter complex self-management challenges and ongoing information needs that traditional health care often fails to meet. Retrieval-augmented generation (RAG) enhances the factual accuracy of large language models (LLMs), but whether personalized RAG is more effective than general RAG remains unclear. Objective This study aimed to assess whether personalized constraint retrieval-augmented generation (C-RAG) is more effective than open-domain general retrieval-augmented generation (O-RAG) and standard care in reducing information needs among patients receiving dialysis and to identify predictors of chatbot engagement. Methods A sequential 2-phase mixed methods study was conducted. Phase 1 evaluated 4 LLMs (Claude 3.5 Sonnet [Anthropic], Gemini 1.5 Flash [Google], CLOVA X [NAVER], and ChatGPT-4o [OpenAI]) using 60 patient-generated questions, assessed by 15 clinical experts. Phase 2 was a 3-arm parallel pilot randomized controlled trial (RCT; October-December 2025) at 3 tertiary hospitals in South Korea. Based on phase 1 findings, 2 chatbot variants were developed using GPT-4o—C-RAG, which incorporates patient-specific clinical parameters, and O-RAG, which uses the same knowledge base without personalization. Participants were randomized 1:1:1 to C-RAG, O-RAG, or standard care. The primary outcome, change in medication, dietary, and diagnostic information needs (5-point Likert scales) from baseline to 4 weeks, was analyzed using ANCOVA. Semistructured interviews with 9 participants explored underlying mechanisms. Results ChatGPT-4o achieved the highest scores in clinical accuracy (mean 3.60, SD 0.95), safety (mean 3.67, SD 1.00), and readability (Flesch-Kincaid Grade Level 9.8, compared with 13.2‐14.5 for other models). In phase 2, 42/45 (93%) enrolled participants completed the study (14 per group). ANCOVA revealed significant group differences in medication ( F 2,38 =4.574, P =.02) and dietary ( F 2,38 =4.232, P =.02) information needs. In Bonferroni-corrected comparisons, C-RAG had lower needs than O-RAG in both domains (observed mean 2.93, SD 1.21 vs mean 3.86, SD 0.95; Cohen d =−0.86, 95% CI −1.63 to −0.08; P =.02 for both). Against control, C-RAG showed medium-to-large but nonsignificant effects (medication Cohen d =−0.64, 95% CI −1.39 to 0.12; dietary Cohen d =−0.54, 95% CI −1.30 to 0.21). Secondary outcomes did not differ between groups. Dietary management emerged as the most frequent query topic, accounting for 29.3% (205/691) of total queries. Participants on dialysis for 1‐3 years showed higher engagement (9/10, 90% engagement for 1‐3 y vs 5/18, 27.8% for ≥4 y; Fisher exact P =.006). Qualitative analysis revealed 5 themes, including dietary management as the primary need and dialysis duration as a moderator of utility. Conclusions In this pilot trial, personalized RAG showed lower information needs than general RAG, with a large effect size (Cohen d =−0.86). The medium-to-large effects against standard care (Cohen d =−0.54 to −0.64) support a fully powered confirmatory trial (estimated 50‐55 participants per group) to evaluate long-term clinical outcomes.

What the data shows

  • No database records a citation yet, as is usual for a paper this new.
  • Ki Won Lee has the highest h-index among the authors (7).

Graph

This paper in the middle, its authors and their most-cited papers below, the papers it cites to the left and the papers citing it to the right. The timeline lays every paper out by year.

Citing paper shares an authorAuthor without a profileSize shows citations, or h-index for authors

Select a node for its details. Drag to pan, scroll to zoom, drag a node to move it.

Start with

Authors (11)

Each author's standing in QAI's scholar index: the works their ORCID iD is attached to, and for anyone whose ORCID record names a lab OpenResearch follows, their arXiv preprints matched by name and a shared co-author. Names are as printed on the paper.

ProfileMost-cited paper

Cited by 0 papers

Who cites this paper, from open citation data, with Crossref's record of each; the 100 newest when there are more.

0 shown
No citing papers for this filter.

References (35)

What it cites, from open citation data, with Crossref's record of each.

Sign in to see 15 more references

A free account opens the rest of the citing papers and references on every paper’s page.